On Thu, Mar 25, 2021 at 06:58:56PM +0100, Ilya Maximets wrote:
> On 3/25/21 5:43 PM, Stefan Hajnoczi wrote:
> > On Thu, Mar 25, 2021 at 12:00:11PM +0100, Ilya Maximets wrote:
> >> On 3/25/21 10:35 AM, Stefan Hajnoczi wrote:
> >>> On Wed, Mar 24, 2021 at 02:11:3
On Wed, Mar 17, 2021 at 09:25:26PM +0100, Ilya Maximets wrote:
Hi,
Some questions to understand the problems that SocketPair Broker solves:
> Even more configuration tricks required in order to share some sockets
> between different containers and not only with the host, e.g. to
> create service c
On Thu, Mar 18, 2021 at 08:47:12PM +0100, Ilya Maximets wrote:
> On 3/18/21 6:52 PM, Stefan Hajnoczi wrote:
> > On Wed, Mar 17, 2021 at 09:25:26PM +0100, Ilya Maximets wrote:
> >> And some housekeeping usually required for applications in case the
> >> socket serve
On Thu, Mar 18, 2021 at 09:14:27PM +0100, Ilya Maximets wrote:
> On 3/18/21 8:47 PM, Ilya Maximets wrote:
> > On 3/18/21 6:52 PM, Stefan Hajnoczi wrote:
> >> On Wed, Mar 17, 2021 at 09:25:26PM +0100, Ilya Maximets wrote:
> >> BTW what is the security model of the b
Hi Ilya,
By the way, it's not clear to me why dpdkvhostuser is deprecated. If OVS
is restarted then existing vhost-user connections drop with an error but
QEMU could attempt to reconnect to the UNIX domain socket which the new
OVS instance will set up.
Why is it impossible to reconnect when OVS ow
On Fri, Mar 19, 2021 at 04:37:01PM +0100, Ilya Maximets wrote:
> On 3/19/21 3:16 PM, Stefan Hajnoczi wrote:
> > On Thu, Mar 18, 2021 at 09:14:27PM +0100, Ilya Maximets wrote:
> >> On 3/18/21 8:47 PM, Ilya Maximets wrote:
> >>> On 3/18/21 6:52 PM, Stefan Hajnoczi wrot
On Fri, Mar 19, 2021 at 04:29:21PM +0100, Ilya Maximets wrote:
> On 3/19/21 3:05 PM, Stefan Hajnoczi wrote:
> > On Thu, Mar 18, 2021 at 08:47:12PM +0100, Ilya Maximets wrote:
> >> On 3/18/21 6:52 PM, Stefan Hajnoczi wrote:
> >>> On Wed, Mar 17, 2021 at 09:25:2
On Tue, Mar 23, 2021 at 04:54:57PM -0400, Billy McFall wrote:
> On Tue, Mar 23, 2021 at 3:52 PM Ilya Maximets wrote:
>
> > On 3/23/21 6:57 PM, Adrian Moreno wrote:
> > >
> > >
> > > On 3/19/21 6:21 PM, Stefan Hajnoczi wrote:
> > >> On Fri, Ma
On Wed, Mar 24, 2021 at 02:11:31PM +0100, Ilya Maximets wrote:
> On 3/24/21 1:05 PM, Stefan Hajnoczi wrote:
> > On Tue, Mar 23, 2021 at 04:54:57PM -0400, Billy McFall wrote:
> >> On Tue, Mar 23, 2021 at 3:52 PM Ilya Maximets wrote:
> >>
> >>>
On Wed, Mar 24, 2021 at 02:11:31PM +0100, Ilya Maximets wrote:
> On 3/24/21 1:05 PM, Stefan Hajnoczi wrote:
> > On Tue, Mar 23, 2021 at 04:54:57PM -0400, Billy McFall wrote:
> >> On Tue, Mar 23, 2021 at 3:52 PM Ilya Maximets wrote:
> >>> On 3/23/21 6:57 PM, Adrian M
On Thu, Mar 25, 2021 at 12:00:11PM +0100, Ilya Maximets wrote:
> On 3/25/21 10:35 AM, Stefan Hajnoczi wrote:
> > On Wed, Mar 24, 2021 at 02:11:31PM +0100, Ilya Maximets wrote:
> >> On 3/24/21 1:05 PM, Stefan Hajnoczi wrote:
> >>> On Tue, Mar 23, 2021 at 04:54:
bind.py to support virtio-vhost-user:
4aee6f653 usertools: add virtio-vhost-user devices to dpdk-devbind.py
Finally, vhost-scsi seems broken to me so two workarounds were needed so it can
be tested again:
b9d17bfaf WORKAROUND revert virtio-net mq vring deletion
cadb25e7d WORKAROUND
vhost-user code.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/Makefile| 2 +-
lib/librte_vhost/vhost.h | 35 +++-
lib/librte_vhost/trans_af_unix.c | 49
lib/librte_vhost/vhost.c | 1 +
4 files changed
ut this is a temporary
measure and will be cleaned up in later patches. By simply moving code
unmodified in this patch it will be easier to review the actual
refactoring that follows.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 65 +
lib/librte_vhost/socket.c
.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 9 +
lib/librte_vhost/socket.c| 6 --
lib/librte_vhost/trans_af_unix.c | 5 +
3 files changed, 18 insertions(+), 2 deletions(-)
diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h
index
Introduce a vhost_transport_ops->socket_start() interface so the
transport can begin establishing vhost-user connections. This is part
of the AF_UNIX transport refactoring and removes AF_UNIX code from
vhost.h and socket.c.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhos
socket_cleanup() to
vhost_transport_ops.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 31 +-
lib/librte_vhost/socket.c| 10 ++--
lib/librte_vhost/trans_af_unix.c | 55
3 files changed, 72 insertions
The socket reconnection code is highly specific to AF_UNIX, so move the
remaining pieces of it into trans_af_unix.c.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 10 +++---
lib/librte_vhost/socket.c| 4
lib/librte_vhost/trans_af_unix.c | 9
.
Move struct vhost_user_connection to trans_af_unix.c and
conn_list/conn_mutex into struct af_unix_socket.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 19 ++---
lib/librte_vhost/socket.c| 36 ++--
lib/librte_vhost/trans_af_unix.c | 60
The fdset is used by the AF_UNIX transport code but other transports may
not need it. Move it to trans_af_unix.c and then make struct vhost_user
private again since nothing outside socket.c needs it.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 15 ---
lib
turn towards vhost_user.c and its mixture of
vhost-user protocol processing and socket I/O. The socket I/O will be
moved into trans_af_unix.c so that other transports can be added that
don't use file descriptors.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 2 +-
lib/li
.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 27 --
lib/librte_vhost/trans_af_unix.c | 41 +++-
lib/librte_vhost/vhost.c | 3 ++-
lib/librte_vhost/vhost_user.c| 18 +-
4 files changed, 68
The core vhost-user protocol code should not do socket I/O because the
details are transport-specific. Move code to send and receive
vhost-user messages into trans_af_unix.c.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 26
lib/librte_vhost/vhost_user.h
these structs together.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 11 +-
lib/librte_vhost/trans_af_unix.c | 44 +---
lib/librte_vhost/vhost.c | 10 -
3 files changed, 34 insertions(+), 31 deletions(-)
diff
lable from vhost_user.c.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 17 +++
lib/librte_vhost/vhost_user.h| 3 ++
lib/librte_vhost/trans_af_unix.c | 78 +
lib/librte_vhost/vhost_user.c| 95 ++---
to librte_vhost without exporting all
private vhost.h symbols.
Signed-off-by: Stefan Hajnoczi
---
Is there a better way of giving a PCI driver access to librte_vhost
transport (not a public API that apps should use)?
---
drivers/Makefile| 2 ++
{lib => driv
it. Or perhaps we could backport SPDK's
lib/virtio. I don't have time to do either right now so I've just
copied the code, removed virtio-net and ethdev parts, and renamed
symbols to avoid link errors.
Signed-off-by: Stefan Hajnoczi
---
drivers/librte_vhost/Makefile |
Currently the only way of determining a struct vhost_virtqueue's index
is to search struct virtio_net->virtqueue[] for its address. Stash the
index in struct vhost_virtqueue so we won't have to search the array.
This new field will be used by virtio-vhost-user.
Signed-off-by: St
Extend the API to support the virtio-vhost-user transport
as an alternative to the AF_UNIX transport. The caller provides a PCI
DomBDF address:
rte_vhost_driver_register(":00:04.0",
RTE_VHOST_USER_VIRTIO_TRANSPORT);
Signed-off-by: Stefan Hajnoczi
--
intended use case is for running DPDK inside a guest. Other guests
can communicate via DPDK's "vhost" vdev driver.
For more information on virtio-vhost-user, see
https://wiki.qemu.org/Features/VirtioVhostUser.
Signed-off-by: Stefan Hajnoczi
---
drivers/librte_vhost/Makefile
The new virtio-transport=0|1 argument enables virtio-vhost-user support:
testpmd ... --pci-whitelist :00:04.0 \
--vdev vhost,iface=:00:04.0,virtio-transport=1
Signed-off-by: Stefan Hajnoczi
---
drivers/net/vhost/rte_eth_vhost.c | 13 +
1 file changed, 13
The default filename built into examples/vhost_scsi may not be
convenient. Allow the user to specify the full UNIX domain socket path
on the command-line.
Signed-off-by: Stefan Hajnoczi
---
examples/vhost_scsi/vhost_scsi.c | 93
1 file changed, 75
The new --virtio-vhost-user-pci command-line argument uses
virtio-vhost-user instead of the default AF_UNIX transport.
Signed-off-by: Stefan Hajnoczi
---
examples/vhost_scsi/vhost_scsi.c | 14 +++---
1 file changed, 11 insertions(+), 3 deletions(-)
diff --git a/examples/vhost_scsi
The virtio-vhost-user PCI adapter is not detected in any existing group
of devices supported by dpdk-devbind.py. Add a new "Others" group for
miscellaneous devices like this one.
Signed-off-by: Stefan Hajnoczi
---
usertools/dpdk-devbind.py | 8
1 file changed, 8 insertion
The EVENT_IDX code in DPDK is broken. It's missing the
signalled_used_valid flag that handles the corner cases (startup and
wrapping). Disable it for now.
Signed-off-by: Stefan Hajnoczi
---
examples/vhost_scsi/vhost_scsi.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/examples/vhost
The virtio-net mq vring deletion code should be in virtio_net.c, not in
the generic vhost_user.c code where it breaks non-virtio-net devices.
Signed-off-by: Stefan Hajnoczi
---
drivers/librte_vhost/vhost_user.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/librte_vhost
ned-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 3 +++
lib/librte_vhost/socket.c | 15 +++
lib/librte_vhost/vhost.c | 17 -
lib/librte_vhost/virtio_net.c | 14 ++
4 files changed, 48 insertions(+), 1 deletion(-)
diff --git a/lib/li
2 for
details.
Patch 1 puts the infrastructure in place to distinguish between the built-in
virtio_net.c driver and generic vhost device backend usage.
Patch 2 fixes the regression by handling VIRTIO_NET_F_MQ only when the built-in
virtio_net.c driver is in use.
Stefan Hajnoczi (2):
vhost: add
host net device
backends. Use the new VIRTIO_DEV_BUILTIN_VIRTIO_NET flag to check
whether virtio_net.c is being used.
This fixes examples/vhost_scsi.
Cc: Maxime Coquelin
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost_user.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
d
si so that vring
call works again.
Cc: Changpeng Liu
Cc: Junjie Chen
Signed-off-by: Stefan Hajnoczi
---
examples/vhost_scsi/vhost_scsi.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/examples/vhost_scsi/vhost_scsi.c b/examples/vhost_scsi/vhost_scsi.c
index da01ad378..3cb4383e9 100644
--- a/exa
Input validation is not applied consistently in vhost_user.c. This
suggests that not everyone has the same security model in mind when
working on the code.
Make the security model explicit so that everyone can understand and
follow the same model when modifying the code.
Signed-off-by: Stefan
fragile and it's unlikely that anyone changing the code
would remember this. A security hole can be introduced accidentally.
This patch switches VhostUserMsg.request fields to uint32_t to avoid the
portability and potential security issues.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vh
says.
Stefan Hajnoczi (8):
vhost: add security model documentation to vhost_user.c
vhost: avoid enum fields in VhostUserMsg
vhost: validate untrusted memory.nregions field
vhost: clear out unused SCM_RIGHTS file descriptors
vhost: reject invalid log base mmap_offset values
vhost: fix
Check if memory.nregions is valid right away. This eliminates the
possibility of bugs when memory.nregions is used later on in
vhost_user_set_mem_table().
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost_user.c | 6 ++
1 file changed, 6 insertions(+)
diff --git a/lib/librte_vhost
vhost_user_set_vring_addr() uses the msg->payload.addr union member, not
msg->payload.state. Luckily the offset of the 'index' field is
identical in both structs, so there was never any buggy behavior.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost_user.c | 2 +-
1
Check the virtqueue size constraints so that invalid values don't cause
bugs later on in the code. For example, sometimes the virtqueue size is
stored as unsigned int and sometimes as uint16_t, so bad things happen
if it is ever larger than 65535.
Signed-off-by: Stefan Hajnoczi
---
match.
Refactoring vhost_user.c to pass around and check the number of file
descriptors everywhere would make the code more complex. It is simpler
for read_fd_message() to set unused elements in fds[] to -1. This way
close(-1) is called and no harm is done.
Signed-off-by: Stefan Hajnoczi
--
If the log base mmap_offset is larger than mmap_size then it points
outside the mmap region. We must not write to memory outside the mmap
region, so validate mmap_offset in vhost_user_set_log_base().
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost_user.c | 9 +
1 file changed
If memory_size + mmap_offset overflows then the memory region is bogus.
Do not use the overflowed mmap_size value for mmap().
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost_user.c | 12 +++-
1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/lib/librte_vhost
icate it.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/rte_vhost.h | 15 +++
examples/vhost/virtio_net.c| 11 ++-
examples/vhost_scsi/vhost_scsi.c | 6 +++---
lib/librte_vhost/vhost.c | 27 +++
lib/librte_
v2:
* Add internal vhost_vring_call() helper function [Maxime]
These patches eliminate code duplication for vhost_virtqueue->callfd users by
introducing rte_vhost_vring_call() (public API) and vhost_vring_call()
(librte_vhost-internal API).
Stefan Hajnoczi (2):
vhost: add vhost_vring_c
Extract the callfd eventfd signal operation so virtio_net.c does not
have to repeat it multiple times.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/vhost.h | 12
lib/librte_vhost/virtio_net.c | 23 +++
2 files changed, 15 insertions(+), 20 deletions
icate it.
Signed-off-by: Stefan Hajnoczi
---
lib/librte_vhost/rte_vhost.h | 15 +++
examples/vhost/virtio_net.c| 11 ++-
examples/vhost_scsi/vhost_scsi.c | 6 +++---
lib/librte_vhost/vhost.c | 21 +
lib/librte_
53 matches
Mail list logo
